Self-stimulating sandwich structure with resistive heating honeycomb core for quick thermographic inspection

The paper presents the novel concept of a honeycomb sandwich structure with carbon fibre reinforced plastics face sheets capable of its own internal thermal stimulation for rapid active thermographic inspection, where an adapted electrically powered honeycomb core serves as a heat source. The proposed sandwich structure effectively reduces the equipment necessary for external thermal stimulation, while improving the controllability of the thermal pulse. A new design of the modified aluminum honeycomb core, required for resistance heating with sufficient homogeneity, is proposed. Numerical modeling was used to test the concept’s viability and to predict its efficiency for defect detection. The altered honeycomb was then manufactured and its heating characteristics were measured. This structure was then used as a core in a sandwich specimen with carbon fibre reinforced plastics face sheets, which also contained artificial defects embedded in the face sheets and in the face sheet/core interface. The applicability of the proposed modified honeycomb structure for thermal stimulation for active infrared thermography was experimentally confirmed, demonstrating the ability to visualize even very small defects simulating disbonding, or delamination. The impact of water ingress on the operation of the concept was tested as well. The proposed method has the potential to simplify and expedite the non-destructive inspection of aviation-grade sandwich structures in service via active infrared thermography.


Introduction
Sandwich structures composed of relatively thin face sheets from carbon fibre reinforced plastics (CFRP) bonded to a honeycomb core made of aluminum or aramid paper (Nomex ® , DuPont) have found applications, e.g., in aerospace due to their excellent mechanical properties, such as high bending stiffness, good fatigue resistance and low specific weight [1]. The sandwich structures can be found in modern aircraft both in primary and secondary load-bearing structures [2,3]. However, these sandwich structures are prone to service-induced damages, such as delamination, fibre and matrix cracking, breaking and scratching, core/face sheet interface defects (disbonding), crushing, cracking, or even water or moisture ingress [4][5][6][7][8][9]. The formation of such damages can be caused by a structure overload during flight, improper handling, or after an impact with a foreign object moving at low speed (low-velocity impact), such as debris, hail, or falling tools, which can cause barely visible impact damage (BVID). BVID usually combines the delamination of the face sheet with the core crushing, disbonding, or creation of a core cavity [10][11][12]. The damage is typically more localized than in the case of monolithic laminates [13], but it can still cause a significant degradation of the mechanical properties of the sandwich.
Furthermore, defects such as delamination, porosity, impurities, cracking, crushing, or disbonding can occur even during manufacturing. They all can negatively impact both the mechanical properties and the service life of the sandwich [14,15]. The correct formation of the adhesive bond between the honeycomb core and the face sheet is critical as well [16,17].
To ensure that the sandwich structures meet the high requirements set by the aerospace industry, thorough non-destructive inspections (NDI) both during manufacturing, and, periodically, when in service, are a necessity. One of the NDI methods used in the industry is active infrared thermography (AIT), which relies on an external excitation source of the thermal wave [18,19]. The thermal properties of the intact sandwich structure and the defect structure differ, affecting the flow of the thermal wave. This difference results in a thermal response, which can be detected and measured by a thermal imager.
Depending on the thermal excitation, several categories of AIT can be recognized. Arguably most widespread is optical radiation (optically stimulated thermography-OST), but also ultrasonic infrared thermography (UIT), eddy current thermography (ECT), or active microwave thermography (AMT) are used [20][21][22][23]. Additionally, the duration, intensity, or even the shape of the generated thermal pulse in combination with the data evaluation method can differ. The classical approach is pulsed thermography (PT), the principle of which lies in a simple pulsed heating of the specimen and subsequent measurement of its cooling process. Other methods include, e.g., step heating thermography (SHT), lock-in thermography (LT), or frequency-modulated thermography (FMT) [18,19,[23][24][25][26].
AIT presents an appealing alternative to other NDI methods, as it yields competitive results in terms of probability of defect detection (POD) in sandwich structures, it is a rapid method with the ability to scan an entire surface or its portion at the same time, and it is non-contact without a requirement for any couplant [27]. However, the proportionality of the detectability with the defect's depth, the high cost of equipment, and the high requirements for operator training are often listed as disadvantages [27]. Arguably, the sensitivity and detection rate of AIT depends on the ability of the operator to properly set up the inspection, minimize the possible interference with the outside world, and properly understand and interpret the resulting thermograms.
The heating process is one of the most important factors, which can impact the ability of the AIT to detect defects in the sandwich structure. Generally, to improve the sensitivity of the AIT, the heating must be intensive, homogenous, and well-controllable. In optically stimulated AIT (OST), halogen lamps are typically used. Yet, this method has inherent downsides-it is difficult to precisely control the thermal pulse generation, the intensity of the heating can be relatively low with a significant dependence on the distance between the lamps and the specimen, and the homogeneity of the heating depends on the reflectivity of the heated surface and its shape.
An option, contrasting the external stimulation, is internal thermal stimulation. E.g., research suggests that it is possible to integrate a heating element into the sandwich structure [28]. This points to the possibility of simplifying the inspection setup and even reducing the direct costs of the inspection. However, the weight of the sandwich structure increases. Furthermore, the cost of the sandwich structure can be expected to rise as well, because its manufacturing is more complicated compared to sandwich utilizing conventional honeycomb core. Additionally, the impact on its mechanical properties is yet to be investigated. Authors in [28] aimed at achieving the best quality of the thermal stimulation, as possible, which was reflected in the design of the integrated heating element. Generally, the quality of the heating can impact the resulting detectability, thus improving the homogeneity of the heating can contribute to superior performance of the AIT. In practice, often the smaller defects are not of interest, and the NDI must reveal larger defects than the size stated by the manufacturer (or by convention). Usually, in service, the POD of defects with a diameter larger than 1 in is of concern [29].
Thus, we hypothesize that the necessity for as good homogeneity of the heating as possible to detect such large defects might not be as great as research indicates. Furthermore, regardless of the quality of the heating, if a highly conductive material is present in the sandwich structure (such as the aluminum honeycomb core), it can be expected that the homogeneity of the thermal wave propagation will be low, impacted by the highly conductive core.
If a requirement for homogeneity of the thermal impulse is lowered, it might be possible to utilize a conductive element already present in the considered sandwich structures-the (aluminum) honeycomb core. In accordance with Joule's law, when a current passes through a conductor, heat is generated, and the sandwich structure can thus be thermally stimulated in a well-controlled manner.
As most of the aviation-grade aluminum honeycomb structures utilize dense hexagonal cells arrangement (e.g., 5 × 5 mm [30]), the heating quality should be sufficient for the detection of defects larger than 1 in.
This type of thermal stimulation also combines the benefits of internal stimulation (reduced equipment costs, normalized thermal stimulation) with inherent advantages-e.g., only a small alteration of the sandwich structure is required, limiting both weight and cost increase. Thus, both AIT inspection setup time and operator level of training could be decreased.
However, the application of this method is limited to the sandwich structures composed of a conductive honeycomb core. Furthermore, the inhomogeneous thermal stimulation impacts the resolution and sensitivity of the method. Additionally, the method could be susceptible to, e.g., water ingress, potentially shorting the honeycomb core.
In this paper, we explore the concept of utilizing the honeycomb core for purposes of thermal stimulation and verify the viability of the concept by detecting the artificial defects simulating disbonding and delamination. Additionally, the structure was exposed to a water ingress, to verify the behavior of the proposed concept with water present in the honeycomb cells, possibly shorting the heating system.
The paper further presents an evolution of the idea, where a conventional honeycomb was first used to determine its heating quality. As the heating was irregular, with some sections of the honeycomb overheating and some not heating at all, it was decided to propose a novel design of the honeycomb core, reflecting the necessary alterations to improve the quality of heating for purposes of AIT.
Thermal stimulation of the specimen with an altered honeycomb core The principle of operation is based on Joule's law-as the current passes through the conducting honeycomb core, heat is generated. This heat is then conducted to the adhesive layer and then to the face sheets of the sandwich structure. Neither conduction or convection by the air inside the cells of the honeycomb structure, nor the radiation of the cell walls to their insides can be omitted. The surface temperature of the sandwich structure is monitored by an infrared imager. If defects are present in the face sheets (e.g., delamination), or in the face sheet/adhesive/core interface (disbonding), the propagation of the thermal wave generated by the structure is impacted. As a result, the temperature field captured by the imager is non-homogenous (see figure 1).
Normally, with OST, based on the position of the thermal imager and the excitation source relative to the tested specimen, two configurations of AIT can be recognized: reflection and transmission [25,31]. However, due to the nature of the thermal excitation, a transmission configuration is effectively used with internal stimulation.
The honeycomb structure needs to be manufactured from an electrically conductive material, e.g., aluminum or its alloys. It is also possible to use a honeycomb structure manufactured from an electrically nonconductive material such as Nomex ® coated with an electrically conductive layer.
To be viable for the application of AIT NDI in practice, the sandwich structure equipped with this type of honeycomb structure must be able to rapidly generate thermal stimulation of sufficient homogeneity. The prerequisite for optimal heating is a homogenous heating of the honeycomb walls.

Design alteration of the conductive honeycomb structure
Critical factors, determining the quality of the heat source, are arguably the homogeneity and the intensity of the Joule heat, generated by the passing DC through the honeycomb structure. Preliminary tests showed that common industry-used aluminum core has an inadequate homogeneity of the thermal excitation (see figure 2). Even after connecting the ends of the honeycomb core via electrical conductors (i.e., parallel connection), entire sections of the honeycomb are not heated. Locations around leading wire coupling are, on the other hand, superheated. The unheated sections are not suitable for AIT, as at least some temperature difference is necessary for the method to work. Furthermore, such a difference in temperature causes additional thermal stresses, creating unwanted loading of the structure during the inspection.
It was thus decided that it is necessary to modify the honeycomb structure to assure the even distribution of the electrical current flowing through the structure. The solution was to divide the honeycomb structure into elements, which are electrically isolated from each other by, e.g., a layer of adhesive. Two methods of electrical connection between these elements are then possible-in series or in parallel, as seen in figure 3. The presented honeycomb structure deviates from commonly used hexagonal honeycomb cells, as a square honeycomb is easier to manufacture. The results presented in the manuscript are thus all assuming the square honeycomb.
Either connection type yields advantages and has some disadvantages. The series connection has an advantage in the possibility of a simple coupling of the elements at their ends. If the elements are connected in parallel, an additional conductor is required at the sides of the honeycomb, increasing the weight of the sandwich structure, and complicating the core manufacturing process. The series circuit leads to the summation of the resistance of all elements, which, due to the inherent low resistance of the aluminum, is an advantage. The significant advantage of the parallel circuit is the inherent tolerance to damage-if single or multiple elements lose the ability to conduct current, the system can still operate at limited functionality. Both connection types  have a problem with the internal resistance of the elements used in the honeycomb structure. If the resistance between elements differs, e.g., due to their different length or heights, the Joule heating generated by these elements differs as well. This problem can be addressed by a modification of the thickness of the wall of the elements to achieve equal resistance of all elements. However, this can be impractical. A more sophisticated processing method, accounting for different resistance between elements, should be used for proper assessment of the results of the AIT.
It can be expected that the introduction of the non-conductive material into the sandwich increases the weight of the entire structure and can negatively impact its mechanical properties. However, proper electrical insulation is a necessity both between the face sheets and the honeycomb elements, otherwise an electrical short can be expected. The impact on the mechanical properties will likely depend on the material of the insulator, as foam-like insulators in between honeycomb segments can negatively affect, e.g., the honeycomb stability. The width of the gap caused by the insulator can correlate with the homogeneity of the thermal insulation, where one can expect that the increase in the width will lead to more homogenous heating, as it will effectively cause a decrease in cell sizing. However, this alteration, while improving the heating uniformity, will severely affect the shape of the honeycomb, causing a change in its mechanical properties. Therefore, a minimal layer of insulator should be used.
The impact of the insulator between the honeycomb and the face sheets likely depends on the material of the insulator. Utilization of materials, which are thermally conductive and electrically non-conductive [32,33], will likely yield the best results. The thermally conductive material of the insulator should spread the thermal wave generated by the aluminum honeycomb, increasing the homogeneity of the resulting heating.

Experimental verification of thermal properties of the altered honeycomb core
For the experimental verification of the suitability of the proposed concept of internal thermal stimulation, a series connection of the elements was chosen. The honeycomb was designed, manufactured, and tested for both the intensity and homogeneity of the heating. Ideally, the honeycomb structure should have been similar to the honeycomb structures commonly used in the aerospace industry, especially regarding the size and shape of its cells, its height, and wall thickness. However, manufacturing difficulties prevented the creation of such a honeycomb core. A simple, manual shaping of elements from aluminum sheets with subsequent curing with an adhesive to form the honeycomb core was used (corrugation process) [34]. The chosen shaping method stipulated a few simplifications. Instead of more commonly used hexagonal cells, square cells of relatively large size were used (10.0 × 10.8 mm). The elements of the honeycomb structure were manufactured from cold-rolled aluminum tape with a thickness of 0.15 mm and a height of 12 mm. The thickness is larger than typically used in aviation-grade honeycomb structures [30]. However, the manufacturing process required an increased thickness to ensure the stability of the bent aluminum sheet. In total, 16 elements were manufactured. These were bonded together with a double-sided foam tape with a thickness of 0.5 mm. The dimensions of the honeycomb structure and its final shape are shown in figures 4 and 5.
To achieve a series connection between elements, a conductive connection of the ends of elements had to be assured. For this purpose, a conductive steel sleeve was welded to the elements. The measured resistance of a single element of the honeycomb structure is 3.5 mΩ, and the total resistance of the honeycomb structure is then 54.4 mΩ.
Infrared imager Micro-Epsilon TIM 450 working in the range of 7.5-13 μm and equipped with a 38°× 29°l ens was used in experiments. The optical resolution of the imager is 382 × 288 pixels, the accuracy of the system is ±2% and the temperature resolution is 0.04 K (NETD). For recording, a sampling speed of 10 fps was used. The imager was controlled by the TIM Connect software. The honeycomb structure was connected to the DC power supply Manson HCS-3600 1-16 V/0-60 A. Defined electrical pulses were generated via an electromechanical relay. The power supply was connected to the honeycomb structure by alligator clips capable of sustaining a current of up to 40 A. The schematics, explaining the experimental setup is shown in figure 6.
To determine the heating characteristics of the honeycomb structure after it was subjected to electrical current, the honeycomb structure was placed on the PMI foam bed. The glossy honeycomb with low emissivity  was spray-coated with a ThermaSpray 800 with a known emissivity of 0.96. The imager captured the time course of the heating of the honeycomb structure subjected to different levels of the input voltage, measured at the connecting points of the alligator clips. The tested voltage varied from 0.7 V to 2.0 V. The lower limit was arbitrarily set as to achieve relatively slow heating-a difference of 10°C was achieved in about 18 s. The upper limit was set based on the speed of the heating, where the difference of 10°C was reached already in 2 s, which can be considered as rapid. Furthermore, with the increase in the voltage, the current in the honeycomb increases rapidly. At 2.0 V, the measured current is already 37.5 A, which is close to the 40 A limit of the alligator clips used. The power input reached up to about 75 W. After that, both speed of the heating and the homogeneity could be assessed. For this purpose, measurement areas of 2 × 2 pixels were defined in the TIM Connect software (see figure 7). In these areas, the program calculates a mean value of the measured temperature for each area. A total mean temperature was calculated out of these areas. The time course of the increase of this mean temperature during the heating for all tested input voltages is shown in figure 8 top. As the honeycomb heats, its internal resistance increases, the current drops, and the power input is thus not constant. Yet, considering the relatively small change in the honeycomb temperature, this dependence can be omitted. Figure 7 also shows the homogeneity of the heating of the honeycomb core. The thermogram was captured during 2.0 V heating 2 s after its start. To assess the homogeneity, a variation coefficient was calculated in all measurement areas from the increase of the temperature relative to a reference ambient temperature. The dependency of the coefficient of variation on the temperature increase for a pulse of 2.0 V is shown in figure 8 bottom. The homogeneity of the heating of the honeycomb core can be considered sufficient.

Numerical modeling
To assess the viability of the concept of internal thermal simulation for active infrared inspection, a numerical analysis in ANSYS Workbench 2021R2, Transient Thermal Analysis, was performed. In the simplified numerical model used, the heat transfer from the honeycomb to the sandwich structure was limited only to conduction. Before the model could be simplified, a complete model of a section of the sandwich was tested. Then, the model was step by step simplified, omitting radiation and convection, and in the end, even conduction in the air gaps inside the honeycomb structure, while comparing the results with the original data. Omitting the thermal wave propagation inside of honeycomb cells caused a measurable, but marginal degradation in the homogeneity of the surface heating. As it greatly improved the speed of the simulation, while still providing useful results, it was decided to accept this model reduction. Naturally, the conduction could not be omitted in the simulated defects, and the air gap inside the defect had to be modeled. Furthermore, the cooling of the outer sides of the face sheets was modeled (stagnating air). Radiation of the face sheets to their surroundings is again omitted. The model geometry and its division into bodies is shown in figure 9, and it is designed based on the specimen used in experiments (see Practical verification of the concept).
The honeycomb body (position 4) with its geometry matches a honeycomb structure used for experimental verification of the proposed concept. The thermal conductivity coefficient k of aluminum is equal to 237 W.m −1 .K −1 at 20°C. The upper face sheet body with dimensions of 100 × 135 mm and thickness of 0.5 mm contains three defects placed 0.25 mm under its surface (position 1). These defects correspond to the milled defects in the specimen used for experimental verification. The inside volume of these defects is filled with additional bodies, simulating the layer of air (position 2, thermal conductivity of 0.026 W.m −1 .K −1 ), to achieve the transfer of heat in these gaps at least by conduction. Both face sheets are made of a carbon-epoxy composite. Their thermal properties are thus highly anisotropic (thermal conductivity of 2.5 W.m −1 .K −1 in face sheets and 0.5 W.m −1 .K −1 through thickness was set [35,36]). The dimensionally identical lower face sheet is at position 8. The adhesive (epoxy resin), interfacing the honeycomb core and both face sheets, is created by two bodies (position 3 and position 6) with a thickness of 0.3 mm and a thermal conductivity of 0.3 W.m −1 .K −1 [37]). This rather significant thickness was selected to simplify the modeling of a much more complex actual interface between the honeycomb core and face sheets where the adhesive layer must rise into the honeycomb cells and create fillets [16,17]. The adhesive body was also used to model disbonding defect, which was achieved by cuts both in the upper and lower body. These cuts are again filled with bodies simulating the presence of the air (positions 4 and 7).
The mesh had a node size of 2.10 -3 m, mechanical physics preference, 326592 nodes, and 201495 elements. The mesh generation was software-controlled.
Pulsed heating of the honeycomb body was modeled by assigning the heat flow of 70 W to the entire surface of the honeycomb body for a duration of 4 s. Considering relatively small temperature change, and thus a small change in the resistance of the aluminum, this value was kept constant during the entire process of heating. The heat flow was chosen to correspond to a maximal 5.4°C heating during the experiments, achieved during the 2.0 V heating lasting for 4 s.
At the pre-selected time (5 s after the heating ended), the distribution of the temperature on the outer side of the upper face sheet was shown in the ANSYS. The 'Rainbow' palette was chosen, to improve the visibility of the temperature fields. An automatic color palette assignment based on the current min/max temperatures in the thermogram was used.
It is possible to state that all defects located at the upper face sheet are detectable (see figure 10). However, due to the inhomogeneity of the heating, caused by the relatively considerable sizing of the honeycomb cells, their shape is difficult to recognize. The smallest disbonding defect of a triangular shape and the narrowest milled defect are both at the edge of visibility. Disbonding defect at the lower face sheet is also visible-thanks to the  disruption of the contact between the lower face sheet and the honeycomb core, the thermal wave is not propagated to the lower face sheet and the honeycomb core increases its temperature at the defect location.
Numerical modeling was used to determine the impact of the honeycomb cell size on the visibility of the defects. A model, with a honeycomb cell size of 5 × 5 mm, was run through the simulation with the parameters kept the same as with the 10 × 10 mm honeycomb cell size-including the assignment of the heat flow of 70 W to the entire surface of the honeycomb walls for the duration of 4 s. The mesh in this case had 621956 nodes and 355515 elements. The result of the simulation is shown in figure 11. The homogeneity of the temperature distribution on the surface is considerably improved, which further improves the visibility of the defects. It is also possible to better detect defect shapes and sizes, which is consistent with the heating homogeneity requirements for active thermography inspection.

Practical verification of the concept
Based on the promising results obtained by the numerical modeling, it was decided to verify the simulation by the measurement of the specimen containing artificial defects. For experimental verification, an altered honeycomb core with elements connected in series was integrated into the sandwich structure with artificial defects. Both outer face sheets made of carbon/epoxy laminate had dimensions of 100 × 130 mm and a thickness of 0.5 mm. The face sheets were cured to the honeycomb core by a layer of two-component epoxy adhesive. To the upper face sheet, three grooves with a length of 25 mm and with different thicknesses (2, 6, and 12 mm) were milled. These grooves were used to simulate the delamination at the inner side of the face sheet ( figure 12, positions 1, 2, and 3).
An additional type of artificial defect created on the upper face sheet is disbonding. Disbonding of the upper face sheet and honeycomb core was simulated by three triangular-shaped defects with purposely removed adhesive (positions 4, 5, and 6). Disbonding of the lower face sheet and the honeycomb core was simulated by an oval-shaped defect (position 7). At the end of these experiments, the area outside of these defects was used for the simulation of water ingress to the core of the sandwich structure position 8. For practical verification of the concept, the identical setup was used as for the experimental verification of the properties of the adapted honeycomb structure, i.e., thermal imager TIM450 and a power supply Manson HCS-3600. The specimen containing artificial defects was placed in the field of view of the infrared imager in a horizontal position ( figure 13). The course of heating and cooling of the specimen was captured by an infrared imager at 10 fps, which seemed sufficient considering the relatively slow thermal process. The captured data were processed and evaluated in TIM Connect software and Excel/MATLAB. The ambient temperature was 26°C. The face sheets were spray-coated by ThermaSpray 800 with an emissivity of 0.96. A series of measurements with pulses varying both in duration and voltage level was performed, controlling the amount of thermal energy generated by the honeycomb core. The parameters used for these measurements copied the voltage levels tested with the honeycomb (see Experimental verification of thermal properties of the altered honeycomb core), ranging from 0.7 to 2.0 V. However, the pulse length needed to be increased to account for the additional thermal mass in the sandwich caused by the presence of face sheets. Based on these measurements, measurements of the honeycomb structure, and the numerical modeling, it was decided to use the pulse of voltage level equal to 2.0 V with a duration of 4 s.
The TIM Connect software was also used to capture static thermograms at various time points after sample heating was completed. The 'Iron' color palette was used with the automatic generation based on the actual temperature range in the thermogram. The visibility of selected defects on raw thermograms is apparent in figure 14.
All defects are visible, with the smallest artificial defects being at the limit of detectability. The visibility of the defects is caused mainly thanks to the decrease in the temperature at the location of the honeycomb walls, rather than by the air gaps in the middle of the honeycomb cells. Relatively large dimensions of the honeycomb cells lead to a difficult defect shape definition. Subjectively, defects seem to be best visible after longer time after the heating stopped. This is caused by the faster cooling of the honeycomb walls, improving the temperature homogeneity of the intact surroundings. The disbonding defect at the lower face sheet becomes visible after a longer time.
Data evaluation using principal component thermography (PCT) was performed to further assess the performance of the proposed method of thermal heating. The captured thermograms were first manually cropped to remove the surroundings and focus on the honeycomb structure. Furthermore, the data were arranged in a 3D matrix, where a z-axis represented a temporal change of spatial dependencies. Along this z-axis, a wavelet transformation was performed to reduce the noise in the thermograms. The next step was data normalization, where each thermogram was normalized to its mean value. Then, the data were rearranged-for each thermogram, after the end of its first row, the second row was added, etc These row vectors were then transposed and a 2D matrix A was created. In this matrix, columns represented a temporal change, whereas rows represented a spatial change. An economical singular value decomposition (SVD) was calculated:

A USV T =
The empirical orthogonal functions (EOF) of matrix U represent the spatial variation of the defects of the test sample, where the first couple of EOFs typically represent most of the variation of the collected data. The matrix S is a diagonal matrix with arranged singular values. The columns of matrix V represent the principal components. Selected EOFs of the matrix U can be used to reconstruct the image data, with enhanced information about the defect (see figure 15).
Measurements focusing on the effectiveness of the proposed concept to detect water ingress into the honeycomb core of the sandwich structure were performed as well. Additionally, these measurements were designed to test if the presence of water in a non-isolated honeycomb causes a short circuit and subsequent loss of functionality. The ingress of salt water into the honeycomb core was simulated, where a 3% solution of sodium chloride was injected through a drilled hole of a diameter of 0.8 mm through the lower face sheet directly into the center of the honeycomb cell. Two volumes of the solution were tested-0.5 ml, corresponding roughly to 40% of the cell volume, and 1.0 ml, which corresponded roughly to 80% of the cell volume. The specimen was placed horizontally, to simulate the worst-case scenario for the conventional reflective OST technique [7,8]. The heating pulse with a voltage level of 2 V and with a duration of 4 s was used again.
Tests proved that the functionality of the system was not lost, and all segments still conducted electrical current. The homogeneity of the heating seemed virtually unchanged by the presence of the water. However, both water ingress volumes were detectable. Figure 16 shows clearly detectable water ingress defect with a water volume of 0.5 ml. Thermogram was captured at the time of the best subjective visibility-10 s after the heating ended. For comparison, figure 16 also shows a thermogram captured by OST, with external heating provided by a halogen lamp with a power of 1000 W and a pulse duration of 2 s.

Conclusion
The paper proposes a novel design of the self-heating sandwich structure for purposes of active thermography inspection, where the heating is provided directly by the conductive honeycomb core of the sandwich. Both numerical modeling and experiments indicated the viability of the concept for defect detection.  The intensity and the homogeneity of heating of the standard honeycomb core were tested with unsatisfactory results. An altered honeycomb design was thus proposed, improving both parameters in initial tests, with performance sufficient for the internal excitation of the sandwich structures.
Numerical modeling was used to validate the hypothesis and to determine key performance parameters and their impact on the performance of the concept. It was found that while it might be possible to detect defects of set sizes, the honeycomb cell size has a major impact on the homogeneity of the heating and thus a direct impact on the quality of the proposed method of thermography inspection. Based on the results of the numerical analysis, a less favorable design was chosen for the practical validation, mainly due to manufacturing issues. Nonetheless, this design should still be sufficient for defect detection.
Again, as a result of numerical analysis and practical experiments, the highest defect detectability was achieved with a relatively intensive pulse with a duration of only a few seconds long. To assess the defect detectability, the contrast method relying on the assignment of a proper color palette to the thermograms was used. The SVD algorithm for PCT further enhanced the measured results.
The proposed Internal thermal stimulation shows promising results. The visibility of the artificial defect in the upper face sheet and in the interface of the upper face sheet and the honeycomb core is generally lower than with OST. Yet, the method allows for the detection of defects also in the lower face sheet, which is challenging with the external excitation in the reflection configuration.
The proposed concept has the potential to simplify and quicken the NDI by AIT, e.g., in aviation practice. The electrical power to the sandwich could be provided by the aircraft systems or, with suitable connectors, from the external power supply during maintenance. Altered honeycomb core could be used in the entire sandwich structure, or only in critical locations. The sandwich structure does not contain any significant foreign object, such as an integrated heating element presented in [28]. However, it is necessary to integrate an electrical insulator between the walls of the honeycomb structure. The subsequent increase in the weight of the entire structure could be kept minimal, but the impact on the mechanical properties is yet to be tested. As the structure of the honeycomb remains the same and the adhesive layer stabilizes the honeycomb and connects the walls of the honeycomb, its ability to transfer loads should not be significantly decreased.
The issue is the manufacturing of the modified honeycomb structure from aluminum or its alloys. Both commonly used processes for manufacturing honeycomb cores, i.e., expansion, and corrugation [34] can be theoretically used, but it is necessary to apply a thin, yet sufficiently insulating layer of the adhesive. The formation of the conductive connections of the elements of the honeycomb structure could be managed with an agreeable quality by welding or soldering. Manufacturing complex shape sandwich panels could be even more difficult. Additive manufacturing technologies can be also used to 3D print the honeycomb core, either with a subsequent application of the insulating layer or, if the additive methods allow it, directly 3D print also the insulating layer. However, methods based on the sintering of the metal powder (DMLS, SLM) currently have a significant limitation in the achievable minimal thickness of the printed honeycomb structure walls [38]. Additive technologies can further widen the applications of the proposed concept of the altered honeycomb core -changes to the wall thickness, dimensions, and shape of the honeycomb cells, or the density of the honeycomb can be easily controlled. Filament-based methods (FFF, FDM) can also be used for honeycomb manufacturing, assuming a conductive filament is used.
Another issue that needs to be explored is the robustness of the proposed concept in practice and the possibilities of repair if the sandwich structure is damaged. In the tests presented in this paper, the resistance to water ingress into the honeycomb core was successfully tested.
Modification of the existing panels solely for purposes of NDI by AIT can be debatable, considering the issues with manufacturing, and the expected increase in the cost. However, the heated honeycomb core can find other applications as well-e.g., deicing, or drying of the honeycomb core after it was subjected to humidity. The electrically conductive honeycomb core could be also used for rapid check of its state by simply measuring its resistance. The increase in resistance or loss of conductivity could point to damage to the honeycomb core. Additional applications of the conductive honeycomb core in smart structures can be found as well.
Further research in this area will be focused on the identification of the impact of parameters of the honeycomb structure on the results of the AIT inspection, an application of sophisticated data post-processing in combination with modulated heating, and an exploration of manufacturing methods of the altered honeycomb structure.